Scene classification is an active research area in the remote sensing (RS) domain. Some categories of RS scenes, such as medium residential and dense residential scenes, would contain the same type of geographical objects but have various spatial distributions among these objects. The adjacency and disjointness relationships among geographical objects are normally neglected by existing RS scene classification methods using convolutional neural networks (CNNs). In this study, a multi-output network (MopNet) combining a graph neural network (GNN) and a CNN is proposed for RS scene classification with a joint loss. In a candidate RS image for scene classification, superpixel regions are constructed through image segmentation and are represented as graph nodes, while graph edges between nodes are created according to the spatial adjacency among corresponding superpixel regions. A training strategy of a jointly learning CNN and GNN is adopted in the MopNet. Through the message propagation mechanism of MopNet, spatial and topological relationships imbedded in the edges of graphs are employed. The parameters of the CNN and GNN in MopNet are updated simultaneously with the guidance of a joint loss via the backpropagation mechanism. Experimental results on the OPTIMAL-31 and aerial image dataset (AID) datasets show that the proposed MopNet combining a graph convolutional network (GCN) or graph attention network (GAT) and ResNet50 achieves state-of-the-art accuracy. The overall accuracy obtained on OPTIMAL-31 is 96.06% and those on AID are 95.53% and 97.11% under training ratios of 20% and 50%, respectively. Spatial and topological relationships imbedded in RS images are helpful for improving the performance of scene classification.

Natural hay-harvesting grasslands are essential for sustaining livestock through winters in semi-arid temperate steppes. However, limited historical records and the absence of systematic spatial data hinder regional management. Moreover, although threshold-based and machine learning approaches have demonstrated effectiveness in monitoring mowing in Western Europe, their applicability in Eurasian semi-arid steppes remains uncertain. Therefore, we aimed to develop an approach for mapping natural hay-harvesting grasslands in semi-arid temperate steppes. Specifically, based on optical satellite time-series data, we employed a time-series classification architecture, Light Inception with Boosting Technique (LITE), to identify hay-harvesting grasslands. Results of our experiments, conducted in a typical region in Inner Mongolia, demonstrated that our approach can generate high-quality hay-harvesting grassland maps, with the testing accuracy of 92.06% (F1-score) and 90.65% (Overall accuracy). Furthermore, we applied the Gradient-weighted Class Activation Mapping (Grad-CAM) technique to interpret the decision-making process of the deep learning model. Our findings showed that LITE concentrated intensively on time steps surrounding mowing dates, underscoring its superiority in mapping the geospatial extent of hay-harvesting grasslands and its potential for extracting temporal information related to hay harvesting. This study offers valuable insights for developing fine-scale hay-harvesting grassland inventories in semi-arid temperate steppe, an area long overlooked in existing research, thereby supporting the sustainable management of grassland resources, and promoting the advancement of smart herding practices.